NF7 rev2 VTT mod need it so bad...

Quote:

---

Originally posted by uwackme
Well, its a commitment if you take the spreaders off.... warranty is gone. Fan sounds cheaper in the "incase" case. I have good airflow in the case past the rams, so I am set, but I did take the spredders off my CH5 and BH5, Im not sending them back, if one fails, Ill use the remaining good chips for repairs to other sticks....I have access to SMD rework equipment.

Remember, I found those stickers under the Kingston spredder.... it DEFINITELY affected one chip, made it fail early....I figure by design, but its all gone now :D

---

Can you put a FSB number on the benefit removing the spreaders gave you?
I will use a fan in either case, spreaders or not... but anything that gives me a gain is worth it.
If removing them does help a noticable amount it means the spreader design is obviously flawed and I may make up my own ones at work.
________
vaporizers

Quote:

---

Originally posted by HaLDoL
on my board (with 3.3V=vdimm and vref=vtt mod) the vtt tracks nicely the vdimm.
Well sort of ... it tracks the vdimm value I get from the 8rdavcore program, but not the real multimeter measured volts.
Eg, I measured a vtt of 1.77V when the vdimm was 3.55V in 8rdavcore (which corresponds to 3.65V measured with the multimeter).
So it tracks, but it undervolts a bit. Still I can get 265FSB with the vtt-vref mod before my mem craps out. With better mem, I should be able to do 270.

---

awesome FSB u got there... Congrats:toast:

:(

Did vdimm=3.3v mod,
Did both L12 mods,
Did vtt=vref,
Did vdd booster...


230*8 isn't primestable:confused:


I'll try the tictac bios.. (wich i need to use???)

I have 1*512 twinmos bh-5 sitting 3rd slot..

Dave, taking the spredder/sticker off was good for about 11Mhz increase to the point where memtest failed test #5. It was the chip with the sticker that failed first....like 00000000 was FF000000 data on failure. But at the higher FSB without sticker the failures were distributed across various bytes, meaning all the chips were failing now at the new higher speed.

Personally I think pulling the spredders off, removing the sticker (THIS violates the warranty 100%, they are tamperproof stickers), removing the goop, and re-attaching the spredders with real AS3 ceramique (need to bend the clips in a bit to take up slack from "goop thickness") would actually be the best result. You'd make them actual heatsinks :D

Like I said, they arent EVER being returned, as the individual BH5 chips are now gold for repairs.

Quote:

---

Originally posted by Rub87
:(

Did vdimm=3.3v mod,
Did both L12 mods,
Did vtt=vref,
Did vdd booster...


230*8 isn't primestable:confused:


I'll try the tictac bios.. (wich i need to use???)

I have 1*512 twinmos bh-5 sitting 3rd slot..

---

I would recommend the D22 Alpha1 bios. Its been the best one for my board.

http://s6.invisionfree.com/undercloc...hp?showtopic=9

Ok,

I've tried the bios you sad, 240*10 is memtest stable, as i move the multiplayer up, the lower the max stable fsb is, altough i don't reach the maximum cpu clock (3ghz) 240*12 boot, 240*12.5 not..

I've measured vtt/vbt = vref = 1.65v while vdimm=+3.3v = 3.45v while running memtest-86 , do i need to cut the pcb to do the other vtt mod so i can manually adjust vtt/vbt to 1.71v so it's exactely the half of vdimm? and do i need to remove the vtt/vbt=vref mod then?

Without a fan on the ram i can't touch the module for 5 sec after a few hours memtest-86 @ 240 fsb, without fan i get some minor errors after a while.. so the module gets verry hot also...

Need soem help here, i want to reach 240*12.5 memtest stable and at least 250fsb @ low multi...

Quote:

---

Originally posted by uwackme
Dave, taking the spredder/sticker off was good for about 11Mhz increase to the point where memtest failed test #5. It was the chip with the sticker that failed first....like 00000000 was FF000000 data on failure. But at the higher FSB without sticker the failures were distributed across various bytes, meaning all the chips were failing now at the new higher speed.

Personally I think pulling the spredders off, removing the sticker (THIS violates the warranty 100%, they are tamperproof stickers), removing the goop, and re-attaching the spredders with real AS3 ceramique (need to bend the clips in a bit to take up slack from "goop thickness") would actually be the best result. You'd make them actual heatsinks :D

Like I said, they arent EVER being returned, as the individual BH5 chips are now gold for repairs.

---

Thanks uwackme:)... I'll think bout doing that, jus dont want to kill something now that my rig is almost "finished"!

Nearly had a heart attack last night after lapping my NB and finding my system wouldnt boot! Turned out I jus needed to reseat my CPU :)

PS, who is Dave?


[EDIT]
I've been wondering this for a few days... it looks like the vdimm mosfet has 5v-rail input, 3.3v-rail input, and vdimm output.

For those not able to run high 3.3v rail but still wanting to run high vdimm could we just put a variable resistor between the 5v-rail-input and 3.3v-rail-input to bump up the 3.3v signal?

Would this have adverse side effects like causing the 5v signal input to drop significantly?

Jus an idea...
________
DS7

Heh, Dave is Dave.... no I was answering too places at once and got mixed up :rolleyes:

Um that isnt the 5V rail, it is the active part of the switching regulator circuit, that happens to run at around 5V....

I think you could mod the ciruit to run off 5V instead of 3.3, like DFI has but the hard part is getting a high current connection to the 5V rail. More trouble than it's worth....easier to get a DFI board. If they ever did a re-design, NF7-S v3.0 should use the 5V rail for ram, I think we all agree on that.

solder a wire to +5v on ATX connector.

If you are going to give the 3.3v input of the vdimm mosfet 5v the vdimm wil be way to high i think, then you also will need to put a resistor between the other wire with 5v that controls the mosfet..?

I still have the following problem:

240*10 = memtest stable
200*15= memtest stable
240*12= not memtest stable

Those settings were al runned @ the same vcore (2.1v in bios) and with the same timings..

Does anyone have an idea what could be the solution, I measured +3.3v under memtest-86 = 3.40-3.45v while vtt/vbt= 1.63-1.65v while running memtest?


I did the vref=vtt mod, do i need to remove the vref=vtt mod if i want to do the cut-trace vtt mod?

Greetz

Ok I've started doing some Prime95 tests aswell. I will use the same Vdimm & Vtt for each setting to begin with, and vary Vdd or timings where necessary.

Mods:
166FSB L12
Vtt, Vdimm, Vdd
Lapped NB with Swiftech MCX159 HSF & Arctic Ceramique

Settings:
Vdimm = 3v
Vtt = 1.5v
Vdd = 1.6v-1.9v (measure by AGP latch... may raise to 2.0v if I get enough people saying it's safe )


2x512MB BH5 Dual Channel, Dimm1 & Dimm3:

223FSB 2-2-2-11 1.6Vdd fails Prime-Blend within 1-2 minutes
223FSB 2-2-2-11 1.7Vdd fails Prime-Blend within 1-2 minutes
223FSB 2-2-2-11 1.8Vdd fails Prime-Blend within ~5 minutes
223FSB 2-2-2-11 1.9Vdd fails Prime-Blend within ~10 minutes

222FSB 2-2-2-11 1.6Vdd fails Prime-Blend within 1-2 minutes
222FSB 2-2-2-11 1.7Vdd fails Prime-Blend within 2-5 minutes
222FSB 2-2-2-11 1.8Vdd fails Prime-Blend within 5-10 minutes
222FSB 2-2-2-11 1.9Vdd fails Prime-Blend within 5-10 minutes

221FSB 2-2-2-11 1.6Vdd fails Prime-Blend within 1-5 minutes
221FSB 2-2-2-11 1.7Vdd fails Prime-Blend within 5-10 minutes
221FSB 2-2-2-11 1.8Vdd fails Prime-Blend within ~50 minutes
221FSB 2-2-2-11 1.9Vdd fails Prime-Blend within ~80 minutes

220FSB 2-2-2-11 1.6Vdd passed Prime-Blend

2x512MB BH5 Dual Channel, Dimm2 & Dimm3:

221FSB 2-2-2-11 1.6Vdd passed Prime-Blend
225FSB 2-2-2-11 1.6Vdd passed Prime-Blend 7hrs

227FSB 2-2-2-11 1.6Vdd fails Prime-Blend within ~15 minutes
227FSB 2-2-2-11 1.7Vdd fails Prime-Blend within ~70 minutes
227FSB 2-2-2-11 1.8Vdd fails Prime-Blend within ~80 minutes
227FSB 2-2-2-11 1.9Vdd fails Prime-Blend within 2hrs


2x256MB 43B Dual Channel:
waiting on RAM to arrive...

[EDIT]
an interesting note is that my old unmodded board running maximum vdimm (e.g. ~3vdimm ~1.45vtt) maxed out at 220FSB 2-2-2-11 2x512mb bh5 24hrs prime-blend, 3dmark, etc... definately seems to be more a BIOS issue, but only testing will prove it)

WE NEED CPC
________
Harley-Davidson FLHS

Quote:

---

Originally posted by Rub87
I still have the following problem:

240*10 = memtest stable
200*15= memtest stable
240*12= not memtest stable

Those settings were al runned @ the same vcore (2.1v in bios) and with the same timings..

Does anyone have an idea what could be the solution, I measured +3.3v under memtest-86 = 3.40-3.45v while vtt/vbt= 1.63-1.65v while running memtest?


I did the vref=vtt mod, do i need to remove the vref=vtt mod if i want to do the cut-trace vtt mod?

Greetz

---

Rub...do u have APIC mode disable?

Quote:

---

Originally posted by Rabbi_NZ
I've been wondering this for a few days... it looks like the vdimm mosfet has 5v-rail input, 3.3v-rail input, and vdimm output.

For those not able to run high 3.3v rail but still wanting to run high vdimm could we just put a variable resistor between the 5v-rail-input and 3.3v-rail-input to bump up the 3.3v signal?

Would this have adverse side effects like causing the 5v signal input to drop significantly?

Jus an idea...

---

rabbi...on Vdimm mosfet u will find 3 voltage reading
Vdimm, i/o volt and something look similar like 5V rail

this 5v rail similar reading is mosfet gate voltage

the higher the voltage the lower resistance between i/o and vdimm... the higher vdimm goes

this 5v coming from winbond ic.... it is vdimm stenghth

Quote:

---

Originally posted by tictac
Rub...do u have APIC mode disable?

---

Nope.. I'll give it a try!


I've tried, no gain at all in memtest-86.. even when i disable cpu interface @ 240*12 the crap isn't stable in test 5&6 ..:mad:

I'm sick of yhis mobo...

I've tried D10, D22 and D23 subzero bios, but none of them is stable above 240 fsb at low multi...

I just don't know what I'm doing wrong? haldol get 265 fsb also with BH-5...:rolleyes:

updated Prime95 results... still not finished tho :)

for those running 2x512MB BH5 Dual Channel, I would recommend putting your best stick in Dimm3 and the other stick in Dimm2 ;)
________
vaporizers

Hmmz, and for the ones running 1*512 BH-5?

check out my MemTest results a page or so back... I got significantly better results using Dimm3 (furtherest from CPU socket) ;)
________
Kawasaki KR350

I'm running in that dimm a single 512Mb twinmos BH-5 stick.. I supplied a 120mm fan as cooling because thet module got unhealthy warm after a while running memtest @ 3.4v..:D


But yea I don't know anything i could do to archive higher fsb on this setup..:stick:

Make sure the module isnt overheating... before I put a fan over my modules they would fail memtest after about 30mins due to the temp getting so high.

try messing with timings, slots, vdd, BIOS versions.... and give us some tips when you sort it out :D
________
Suzuki K125

I'm running @ 1.85vdd, could that be to much? or not enough? I'll also check if my waterblock makes good contacts with the nb...

u r running a stick of 512MB ram double side ram is it? i dont think u can go higher than u have now :(

Yes, it's double sided:(

Soon I'll recieve 2*256 BH-6.. let's see what they can do;)

Anyone else need 1.95 VDD to get their there Rev2 C1 to 230 FSB? Mine seems to like voltage...: (

Rub87,
what is the highest your 512MB stick can go in Memtest without errors at 3.4v?
________
VAPORIZER